07 Network Layer Slides

7/25/2019 07 Network Layer Slides

1/68

Internet Technology07. Network Layer

Paul Krzyzanowski

Rutgers University

Spring 2013

1March 17, 2013 2013 Paul Krzyzanowski


2/68

Network Layer

Transport Layer (Layer 4)

Application-to-application communication

Network Layer (Layer 3)

Host-to-host communication

Route The path that a packet takes through the network

Routing

The process of determining the path

Forwarding Transferring a packet from an incoming link to an outgoing link

Router

The device that forwards packets (datagrams)

March 17, 2013 2

Router

2013 Paul Krzyzanowski


3/68

Forwarding vs. Routing

Routing

Responsibility over the path Routing algorithms figure out the path a packet should take

Forwarding

A router consults a forwarding table

Examines data in a packets header & uses the table to determine theoutgoing link for the packet

Routing algorithms configure forwarding tables

Switches vs. Routers

Packet switches: transfer data between links based on link layer data (e.g.,

Ethernet) Routers: transfer data between links based on network layer data (e.g., IP)

March 17, 2013 32013 Paul Krzyzanowski


4/68

Network service models: our wish list

What would we like from a network?

Guaranteed delivery (no loss)

Bounded (maximum) delay

In-order packet delivery

Guaranteed constant or minimum bandwidth

Maximum jitter

Jitter = variation in latency

Endpoint authentication & encrypted delivery



5/68

Network service models: what do we get?

IP gives us none of this

Best-effort = no guarantees on delivery, delay, order

Other network architectures provide some of these items

E.g., ATM (Asynchronous Transfer Mode)

ATM CBR (Constant Bit Rate)

Connection setup specifies bandwidth

Network provides constraints on jitter and packet loss

Network guarantees in-order delivery

ATM ABR (Available Bit Rate)

In-order delivery

Guaranteed minimum bandwidth but higher rates if resources available

Feedback to sender if congestion is present



6/68

Virtual Circuit vs. Datagram Networks

Virtual Circuit (VC) Networks

Connection service at the network layer

All routers in the path are involved in the connection

Datagram Networks

Connectionless service at the network layer

Connection-oriented service provided at the transport layer

Only end systems are involved

Routers are oblivious

IP is a datagram network



7/68

Virtual Circuit Networks

Connection setup

Set up route based on destination address Each router commits resources

Each router builds enters the connectionin its forwarding table

Routers maintain connection state information

Communication

Each packet contains a VC#

Forwarding table determines the next link and VC#

Destination address notneeded on each packet; just the VC#

Teardown

Clear connection from forwarding table on each router

March 17, 2013 7

Incoming

interface

Incoming

VC #

Outgoin

g

Interface

Outgoing

VC #

1 12 2 83

1 9 2 101

2 4 1 151Forwarding Table



8/68

Datagram Networks

Packet identified with the destination address

No setup; routers maintain no state information

Routers

Use the destination address to forward the packet

Forwarding table maps destination address to output link

IP addresses are 32 bits

We cant have a forwarding table with 232(4,294,967,296) entries!

Match a range of addresses by matching a prefix

March 17, 2013 8

Prefix Outgoing

Interface

1000 0000 0000 0110 2

0000 0111 0010 3

0000 0111 1

Forwarding Table= Routing Table

192.168.*.* = Rutgers

15.*.*.* = HP

15.[32-47].*.* = HPlongest prefixmatching rule

Routing Protocols



9/68

The Router



10/68

SystemTasks

RoutingProcessor

Router Architecture

March 17, 2013 10

SwitchFabric

Control Plane

Data Plane

Data Plane: Packet Forwarding

Layer 1: Retime & regenerate signal

Layer 2: Rewrite header and checksum

Layer 3: Look up, queue, decrement TTL,regenerate checksum, forward to output port

Note: a porton a router refers to theinput & output interfaces, not atransport-layer port

Control Plane: Routing & Management

Run routing protocols & maintain routing tables

User command interface Accounting

ICMP

Queue management

Layer 2Interface

Input Port

Layer 2Interface

Input Port

Layer 2 Interface

Output Port

Layer 2 Interface

Output Port



11/68

Router Architecture: line cards

A line card is responsible for I/O on a specific interface

March 17, 2013 11

Layer 2Interface

Input Port

Layer 2Interface

Output Port



12/68

Line Card

Line Card

Line Card

Shared Memory - Conventional

Ports

Function as I/O devices in an OS

Packet arrival

CPU interrupt

Copied to memory

Routing CPU determines route

Copies packet to output port

Limitation

Only one memory read/write at a time CPU & bus can be bottlenecks

March 17, 2013 12

SharedMemory

SharedBus

CPU

CPUMemory

Routing

Table

Line Card



13/68

Shared MemoryDistributed CPUs

CPU & copy of routing table in line cards

Lookup and data copy to outputport done by line card

Limitation

Only one memory

read/write at a time Bus can be a bottleneck

March 17, 2013 13

SharedMemory

SharedBus

CPU

CPUMemory

Routing

TableLine CardCPU

Line Card

CPU

Line Card

CPU

Line Card

CPU



14/68

Non-shared MemoryBus Data Path

No shared memory

Bus used to copy packetsdirectly from oneport to another

Limitation

Shared bus canbe a bottleneck

March 17, 2013 14

SharedBus

CPU

CPUMemory

Routing

TableLine CardCPU

Line Card

CPU

Line Card

CPU

Line Card

CPU



15/68

Non-shared MemoryCrossbar Data Path

NN crossbar switching fabric

One port can move a packet to another portwithout blocking otherports

Multiple switchingfabrics can usedto route packetsto the same port

Verdict

Fastest solution

$$$

March 17, 2013 15

CPU

CPUMemory

RoutingTable

Line Card

CPU

Line Card

CPU

Line CardCPU

Line Card

CPUCrossbar

Switch



16/68

Output Port Queuing

If theres a queue at an output port

A packet scheduler chooses one packet for transmission This can be simple first-come-first-served (FCFS)

or take other factors into account(source, destination, protocol, service level)

If the output port queue is full

We have packet loss

A router can decide which packet to drop

Active Queue Management (AQM) algorithms: decide which packets to

drop



17/68

Input Port Queuing

If packets arrive faster than they could be switched

They need to be queued at input ports If multiple queues have a packet for the same output port

Only one will be switched at a time

The others will be blocked and the packets behind them will be blocked too!

Head-of-line blocking

If the queue overflows We have packet loss



18/68

Head-of-line blocking

March 17, 2013 18

SwitchFabric

Layer 2Interface

Input Port

Layer 2Interface

Input Port

Layer 2 Interface

Output Port

Layer 2 Interface

Output Port

Layer 2Interface

Input Port

Layer 2 Interface

Output Port

If this packet has to wait

Then these packets have to wait



19/68

Internet Protocol



20/68

Internet Protocol: Layer 3IP

March 17, 2013 20

Application

Transport

Network

Data Link

Physical

Internet protocol stack

1

2

3

4

5

6

7

IP Layer Components

1. IP Protocol Addressing

Datagrams Fragmentation Packet forwarding

1. Routing Protocols

1. ICMP Error reporting Signaling



21/68

IP Datagram Structure

20 byte fixed part

Variable-size options

March 17, 2013 21

Version Total Length

Identification (Fragment)

Data

32 bits4 bytes

Options (if header length > 5)

20bytes

Header checksum

Source IP address

0 1 2 3 4 5 6 78 910

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

HeaderLength DSCP ECN

13-bit Fragment offsetFlags

Time to Live Protocol

Destination IP address



22/68

IP Datagram: Version

4-bit identification of the protocol used: 4 = IPv4

March 17, 2013 22


Identification

Data

32 bits4 bytes


20bytes

Header checksum

Source IP address

0 1 2 3 4 5 6 78 910

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31







23/68

IP Datagram: Header Length

4-bit header length (in # of 32-bit words)

IP packets usually have no options, so this is usually 5

March 17, 2013 23


Identification

Data

32 bits4 bytes


20bytes

Header checksum

Source IP address

0 1 2 3 4 5 6 78 910

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31







24/68

IP Datagram: DSCP

Differentiated Services Control Point

Identifies class of service for QoS aware routers (e.g., VoIP)

March 17, 2013 24


Identification

Data

32 bits4 bytes


20bytes

Header checksum

Source IP address

0 1 2 3 4 5 6 78 910

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31







25/68

IP Datagram: ECN

Explicit Congestion Notifications

Routers normally do not inform endpoints of congestion ECN is an optional feature to allow them to do so

March 17, 2013 25


Identification

Data

32 bits4 bytes


20bytes

Header checksum

Source IP address

0 1 2 3 4 5 6 78 910

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31







26/68

IP Datagram: Total Length

16-bit value of the entire datagram (including the 20-byte IP header)

March 17, 2013 26


Identification

Data

32 bits4 bytes


20bytes

Header checksum

Source IP address

0 1 2 3 4 5 6 78 910

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31







27/68

IP Datagram: Fragmentation

Fragmentation

Identification: Identifies fragment of an original datagram Flags: control fragmentation or identify if there are more fragments

Fragment offset: offset of fragment relative to original data

March 17, 2013 27


Identification

Data


20bytes

Header checksum

Source IP address

0 1 2 3 4 5 6 78 910

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31







28/68

IP Datagram: Time-To-Live

Hop countdecremented by 1 each time the datagram hits a router

If TTL == 0, discard the packet Keeps packets from circulating indefinitely (common TTL = 6064)

March 17, 2013 28


Identification

Data

32 bits4 bytes


20bytes

Header checksum

Source IP address

0 1 2 3 4 5 6 78 910

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31







29/68

IP Datagram: Protocol

Identifies the protocol in the data portion

TCP = 6, UDP = 17 IANA assigns the numbers

March 17, 2013 29


Identification

Data

32 bits4 bytes


20bytes

Header checksum

Source IP address

0 1 2 3 4 5 6 78 910

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31







30/68

IP Datagram: Header Checksum

1s complement checksum of the header

Router discards packet if corrupt Must be recalculated by the router since TTL (& maybe options) change

March 17, 2013 30


Identification

Data

32 bits4 bytes


20bytes

Header checksum

Source IP address

0 1 2 3 4 5 6 78 910

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31







31/68

IP Datagram: Source & Destination

Identifies source and destination IP addresses

March 17, 2013 31


Identification

Data

32 bits4 bytes


20bytes

Header checksum

Source IP address

0 1 2 3 4 5 6 78 910

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31







32/68

IP Datagram: Options

Extensions to the headerrarely used

Options include: route to destination, record of route, IP timestamp

March 17, 2013 32


Identification

Data

32 bits4 bytes


20bytes

Header checksum

Source IP address

0 1 2 3 4 5 6 78 910

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31







33/68

IP Fragmentation & Reassembly

Remember MTU (Maximum Transmission Unit)?

Maximum size of payload that a link layer frame can carry This limits the size of an IP datagram (and hence a TCP or UDP segment)

What if a router needs to forward a packet that is larger than thatlinks MTU?

Break up the datagram into two or more fragments

Each fragment is a separate IP datagram

IP layer at the end system needs to reassemblethe fragments beforepassing the data to the transport layer



34/68

IP Fragmentation

When an IP datagram is first created

Sender creates an ID number for each datagram (usually value of a counter) DFbit (Dont Fragment) set to 0: fragmenting is allowed

When a router needs to fragment a datagram

Each fragment contains the same ID #, source address, destination address Fragment offset

Identifies offset of the fragment relative to the original datagram in 8-byte blocks

First datagram Offset = 0

All fragments except for the last one have the MF(More Fragments) bit set

March 17, 2013 34

Identification 13-bit Fragment offsetFlags

DF

MF

0



35/68

IP Fragmentation

Example: send 4,000 byte datagram

20 bytes IP header + 3980 bytes data

Outbound link at router has a 1500-byte MTU

March 17, 2013 35

src=68.36.211.59 dest=128.6.4.24 ID=2222 TTL=60 Sum=xxx Offset=0MF=

0Data = 3980 byteslen=4000

src=68.36.211.59 dest=128.6.4.24 ID=2222 TTL=60 Sum=aaa Offset=0MF=


src=68.36.211.59 dest=128.6.4.24 ID=2222 TTL=60 Sum=bbb Offset=185MF=


src=68.36.211.59 dest=128.6.4.24 ID=2222 TTL=60 Sum=ccc Offset=370MF=


Fragment 1

Fragment 2

Fragment 3

1858=1480

3708=2960No more fragments

Recompute checksumfor each datagram



36/68

IP Reassembly

Identification

Receiver knows a packet is a fragment if MF is 1 and/or Fragment Offset is not 0

Matching & Sequencing

Identification field is used to match fragments from the same datagram

Offsets identify the sequence of fragments

Size of original

When the receiver gets the last fragment (MF==0, Offset != 0)

It knows the size of the datagram ((offset8)+length)

Giving up

If any parts are missing within a time limit, discard the packet

Linux: /proc/sys/net/ipv4/ipfrag_time (default 30 seconds)

Once reassembled, pass to protocol that services this datagram



37/68

IP Addressing



38/68

IP Addressing

IPv4 address: 32 bits expressed in dotted-decimal notation

www.rutgers.edu = 0x80064489= 128.6.68.137

Each interfaceneeds to have an IP address

E.g., each link on a router has an address

If your laptop is connected via Ethernet and 802.11, you have 2 IP addresses Every interface at a router has its own address



39/68

Route Aggregation: Subnets

IP address = 32 bits = 232addresses

But addresses cannot be assigned randomly Otherwise routing tables would have to be 232entries long!

and maintaining them would be a nightmare

Instead, assign groups of adjacent addresses to an organization

Route aggregation = use one prefix to advertise routes to multipledevices or networks

March 17, 2013 39

www.rutgers.edu = 128.6.68.137

All hosts in Rutgers start with 128.6

First 16 bits of the IP address identify a host at Rutgers

Routers need to know how to route to just 128.6instead ofall 65,536 (216) possible addresses



40/68

Subnets

Subnet(= subnetwork = network)

Group of IP addresses sharing a common prefix (n high-order bits) A logical network connected to a router (LAN or collection of LANs)

Rutgers subnet = 128.6.0.0/16

CIDRnotation (Classless Inter-Domain Routing)

A/N: Nmost significant (leftmost) bits of address

March 17, 2013 40

www.rutgers.edu = 128.6.68.137

10000000 00000110 01000100 10001001

Network number Host number

Top 16 bits identify thesubnetwork



41/68

Subnet Mask

A subnet mask (or netmask)

A bit mask with 1s in the network number position Address & netmask strips away host bits

Address & ~netmask strips away network bits

For Rutgers, the netmask is

11111111 11111111 00000000 00000000

255.255.0.0

For a 221.2.1.0/26 network, the netmask is

11111111 11111111 11111111 11000000

255.255.255.192

March 17, 2013 41

6 bitshost26 bitsnetwork

16 bitshost16 bitsnetwork


From


42/68

How are IP addresses assigned?

IP addresses are distributed hierarchically

Internet Assigned Numbers Authority (IANA) at the top IANA is currently run by ICANN

Internet Corporation for Assigned Names and Numbers

March 17, 2013 42

RIPENCCAfriNIC LACNICARINAPNIC

IANA

Regional Internet Registries (RIR)

Allocate blocks of addresses to ISPs

ISP ISP ISP ISP ISPRIR Map

ISP ISP

ISP

Your computer(or Internet gateway)- We will look at NATlater- Permanent (static) or temporary (dynamic)

FromLecture 3:

DNS



43/68

Address allocation: its a hierarchy

March 17, 2013 43

Allocates blocks of IP addresses

Allocates one or more blocks of IP addresses

Organization

Allocates an address or one or more blocks of IP addresses

ISP

ARIN

IANA

ARIN administers 79 blocks of IP addresses

Networks

Allocates one or more blocks of IP addresses

Allocates IP addresses to hosts

ISP

ISP gets 200.23.16.0/20

OrgA

OrgB

OrgC

200.23.16.0/23

200.23.18.0/23

200.23.20.0/23

Routing: everythingto 200.23.16.0/20

goes here

Subnetting, dividing a network intosmaller networks, can be repeated ateach level of the hierarchy



44/68

Special addresses

Network address: all host bits 0

Rarely, if ever, used Rutgers = 128.6.0.0

Limited broadcast address: all bits 1

Broadcast address for this network, the local network.

Datagrams are not forwarded by routers to other networks

Broadcast address: all host bits 1

All hosts on the specified subnet get datagrams sent to this address

Routers may or may not forward broadcasts (no for outside an organization)

Rutgers = 128.6.255.255

Loopback address: 127.0.0.1 = localhost

Communicate with your own device

Uses the loopback network interface



45/68

Host Configuration

How do you assign an address to a host?

Manually, configure the device with its IP address

Subnet mask, so it knows what addresses are local

Gateway: default address for non-local addresses not in a routing table

Router that connects the LAN to another network

Name server addresses(s), so it can look up addresses

Automatically, via the Dynamic Host Configuration Protocol (DHCP)



46/68

Dynamic Host Configuration Protocol

Protocol for client to get an IP address and network parameters

It has to work before the client has a valid address on the network!

Use IP broadcasts

DHCP server must be running on the same network link (LAN) Else each link must run a DHCP Relay Agentthat forwards the request to

a DHCP server



47/68

DHCP: Three mechanisms for allocation

1. Automatic allocation

DHCP assigns an permanent IP address to a client

2. Dynamic allocation

DHCP assigns an IP address to a client for a limited period of time

Allows automatic reuse of an address that is no longer needed by the client

3. Manual allocation

A client IP address is assigned by the network administrator


DHCP Th P l


48/68

DHCP: The Protocol

March 17, 2013 48

Client broadcasts DHCP Discover

Client sends a limited broadcast DHCPDiscover UDP message to port 67

Contains random transaction identifier

Server responds with an offer

Server sends a limited broadcast DHCP

Offer UDP message to port 68

Response contains

Matching transaction identifier

Proposed IP address

Subnet mask

Lease time

Client broadcasts DHCP Request

Sends back a DHCP message with acopy of the parameters

This performs selection(if multipleoffers), confirmation of data, extension oflease

Discover

Offer

Request

Serversends DHCP ACK

Sends configuration parameters,including committed IP address

Client Server

ACK

D-O-R-A


NAT N t k Add T l ti


49/68

NAT: Network Address Translation

Every device on the Internet needs an IP address

Every address has to be unique otherwise, how do you address a host?

IP addresses are not plentiful

Does an organization with 10,000 IP hosts really need 10,000 addresses?




50/68


Private IP address space in the organization

One external IP address

NAT Translation Table

Map source address:port in outgoing IP requests to a unique external address:port

Inverse mapping for incoming requests

A NAT-enabled router looks like a single device with one IP address

March 17, 2013 50

Internal External

Host Port Host Port

192.168.32.4 1200 68.36.210.55 4000

192.168.32.6 1200 68.36.210.55 4001

192.168.32.6 1102 68.36.210.55 4002

192.168.32.11 9000 68.36.210.55 4003

Translation Table in a NAT-Enabled Router

src: 68.36.210.55:4000dest: 74.125.26.103: 80




51/68


NAT requires a router to look at the transport layer!

Source port (outgoing) & destination port (incoming) changes

TCP/UDP checksum recomputed

March 17, 2013 51

Internal External

Host Port Host Port

192.168.32.4 1200 68.36.210.55 4000

192.168.32.6 1200 68.36.210.55 4001

192.168.32.6 1102 68.36.210.55 4002

192.168.32.11 9000 68.36.210.55 4003

Translation Table in a NAT-Enabled Router


NAT P i t Add


52/68

NAT: Private Addresses

We cannot use IP addresses of valid external hosts locally

how will we distinguish local vs. external hosts?

RFC 1918: Address Allocation for Private Internets

Defines unregistered, non-routable addresses for internal networks

March 17, 2013 52

Address Range # addresses CIDR block

10.0.0.010.255.255.255 16,777,216 10.0.0.0/8

172.16.0.0172.31.255.255 1,048,576 172.16.0.0/12

192.168.0.0192.168.255.255 65,536 192.168.0.0/16


NAT i t


53/68

NAT variants

Static NAT

One-to-one mapping between internal and external addresses

Dynamic NAT

Maps an unregistered (internal) IP address to one of several registered IPaddresses

Overloading, or Port Address Translation (PAT)

A form of dynamic NAT that maps multiple internal IP addresses to a singleregistered address by using different ports


Ad t f NAT


54/68

Advantages of NAT

Internal address space can be much larger than the addressesallocated by the ISP

No need to change internal addresses if ISP changes your address

Enhanced security A computer on an external network cannot contact an internal computer Unless the internal computer initiated the communication

But can only contact the computer on that specific port(this is where active mode FTP had problems)



55/68

ICMP


I t t C t l M P t l (ICMP)


56/68

Internet Control Message Protocol (ICMP)

Network-layer protocol to allow hosts & routers to communicatenetwork-related information

ICMP information is carried as IP payload


ICMP Segment Str ct re


57/68

ICMP Segment Structure

Variable-size segment; 8-byte minimum

Type: command or status report ID

Code: status code for the type

Checksum: Checksum from ICMP header & data

Rest of header: depends on type

Error reports contain the IP header & first 8 bytes of original datagrams data

March 17, 2013 57

Type Checksum

Rest of Header (4 bytes)

32 bits4 bytes

8bytes

0 1 2 3 4 5 6 78 910

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

Code

Data (optional)


Some ICMP Message Types


58/68

Some ICMP Message Types

March 17, 2013 58

Type Description

0 Echo reply (ping)

3 Destination unreachable

4 Source quench

5 Redirect message

8 Echo request

9 Router advertisement

10 Router solicitation

11 TTL exceeded

12 Bad IP header

13 Timestamp14 Timestamp reply

17 Address mask request

18 Address mask reply


Ping program


59/68

Ping program

Get a network ping (echo) from a requested host

Test network reachability Measure round-trip time

Optionally specify packet size

Request/response protocol Ping Client

Create socket (AF_INET, SOCK_RAW, IPPROTO_ICMP)

Set IP header fields & ICMP header fields

Send it to a destination via sendto()

Wait for a response from the destination address via recvfrom()


Ping program


60/68

Ping program

Request

Send ICMP type=8 (echo request), code 0 (no options to echo)

Reply

Destination responds back with an ICMP type=0 (echo reply), code=0

March 17, 2013 60

Type = 8 ChecksumCode = 0

Data

Sequence numberIdentifier

Timestamp Data

Associatereplies with

requests

Type = 0 ChecksumCode = 0

Same data

Same sequence numberSame identifier

same timestamp Same data


Ping program


61/68

Ping program

Get a network ping (echo) from a requested host

Test network reachability Measure round-trip time

Optionally specify packet size

Request

Send ICMP type=8 (echo request), code 0 (no options to echo)

Destination responds back with an ICMP type=0 (echo reply), code=0

March 17, 2013 61

Type Checksum

8bytes

0 1 2 3 4 5 6 7 8 910

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

Code

Data

Sequence numberIdentifier

Time stamp Data


Traceroute program


62/68

Traceroute program

Traceroutetrace a route to a specific host

Send a series of UDP segments with a bogus destination port 33434 to 33534 on Linux systems

First IP datagram has TTL=1

Second IP datagram has TTL=2, and so on

Keep a timer for each datagram sent

At a router

When the TTL expires, a router sends an ICMP warning message

Type 11, code 0 = TTL expired

ICMP message includes the name of the router and its IP address

At the final destination The destination sends an ICMP warning message

Type 3 code 3 = Destination port unreachable


IPv6


63/68

IPv6

Weve been rapidly using up IPv4 addresses ever more rapidly

Growth of the web Always-on IP devices

Set-top boxes and phones

Inefficient network allocation

We dealt with it with

NAT

Name-based web hosting

Reallocation of network allocation & subnetting

Those solutions helped a lot but not enough

Were out of IPv4 addresses in parts of the world

IPv6 to the rescue!


Highlights


64/68

Highlights

Huge address space

128-bit addresses: 3.41038

addresses (>7.91028

more than IPv4) Simplified 40-byte header

Longer addresses but far fewer fields

Focus is to simplify routing

Anycast address Allows a datagram to be delivered to one of a group of interfaces

Usually used to identify the nearest host of several hosts

Flow label

Allows related packets that require specific levels of service to be identified

E.g., voice, video

Not well defined yet




65/68


March 17, 2013 65

Version Flow label (20 bits)

Payload length

Data

32 bits4 bytes

40byte

s

Source IP address (128 bits)

0 1 2 3 4 5 6 78 910

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

Traffic class (8 bits)

Hop limit (8 bits)Next header (8 bits)

Destination IP address (128 bits)


IP datagram structure


66/68

IP datagram structure

Version: protocol version (not compatible with IPv4!)

Traffic class: category of service

Flow label: identification tag for related flows

Payload length: # bytes following the 40-byte datagram

Next header: identifies higher-level protocol (e.g., TCP or UDP)

Same as in IPv4

Hop limit: TTL; decremented at each router

Source & destination addresses

Data No fragmentation!

No header checksum! Ethernet does it; so do TCP and UDP


Transitioning


67/68

Transitioning

IPv6 systems can bridge to IPv4 networks

IPv4 addresses are a subset of IPv6 addresses Dual-stack systems

Hosts with both IPv4 and IPv6 network stacks to communicate with bothprotocols

DNS can identify if a given domain is IPv5 capable or not

IPv4 systems cannot communicate with IPv6 systems

Migrating to IPv6 results in a loss of global visibility in the IPv4 network

Initial transition is not visible to end users

Cable modems, set-top boxes, VoIP MTAs

IPv6 access



68/68

The end

07 Network Layer Slides

Documents

Transcript of 07 Network Layer Slides